We are grateful to Francesco Guala for providing a thoughtful reflection on what recent social dilemma experiments can tell us about real-world cooperation and the need for complementary ethnographic, historical approaches. But Guala's contribution is packaged along with what we think is a misunderstanding of our work, an overly pessimistic appraisal of the external validity of experimental results, and a very partial reading of the evidence on costly punishment in small-scale societies.

The core of strong reciprocity is that human cooperation cannot be understood entirely as the result of repeated social interaction and self-interested individual calculation. Instead, people are motivated to cooperate with one another and to punish free-riding by a variety of ethical and other-regarding motives. Guala gets this right. However, he incorrectly believes that strong reciprocity requires punishment to be both very costly and uncoordinated. Punishment is costly when the cost of administering punishment, however small, exceeds the private benefit it creates for the punisher, thus giving rise to a second-order free-rider problem. Mechanisms like conformism, kin selection, or cultural group selection can solve the second-order free-rider problem, but usually only if the cost of punishment is low, either because it is rare (e.g., Boyd et al. Reference Boyd, Gintis, Bowles and Richerson2003; Henrich & Boyd Reference Henrich and Boyd2001) or because it is collectively administered (Boyd et al. Reference Boyd, Gintis and Bowles2010).

Everyday social life, even among strangers, is regulated by many individual acts of uncoordinated punishment. We are all aware of the pain we experience when we are frowned upon in public places among strangers. However, we agree with Guala that more costly forms of punishment in natural settings are usually collective. We capture this in our paper “Coordinated Punishment of Defectors Sustains Cooperation…” (Boyd et al. Reference Boyd, Gintis and Bowles2010), which Guala cites but seems to have misunderstood. In this model, potential punishers signal their willingness to punish, but they punish free-riders only when enough fellow punishers have signaled. When there is no assortment, there are two possible evolutionary equilibria: a population without punishment or cooperation, and a population with a mix of punishers and non-punishers in which most actors cooperate. Mean fitness is higher when punishers are present. When we allow an empirically realistic degree of assortment in the population, punishment may proliferate even when rare; and when it does, it is altruistic.

We developed this model because we share Guala's dissatisfaction with the typical representation of punishment as an individual act rather than something deliberated on by groups and undertaken jointly (but see Ertan et al. Reference Ertan, Page and Putterman2009). Nonetheless, experiments make a major contribution by showing that the predispositions that motivate punishment are common in many populations. We agree with Guala that we need better tests of the external validity of these experimental results. But two kinds of evidence are encouraging.

First, behavior in experiments predicts subjects' cooperation in the field. Brazilian shrimpers use large plastic bucket-like contraptions in which holes are cut to allow the immature shrimp to escape, thereby preserving the stock for future catches. Because they can cut holes of any size, the fishermen face a real-world social dilemma. Large holes represent cooperation with other fishers; small trap holes are a form of defection, and – just as in the Public Goods Game – having small holes is the dominant strategy for a self-interested shrimper. Not surprisingly, those who contributed most in a public goods experiment were also those who cut larger holes in their traps (Fehr & Leibbrandt Reference Fehr and Leibbrandt2011). The effects, controlling for a number of other possible influences on hole size, are substantial.

Additional evidence of external validity comes from a set of experiments and field studies with 49 groups of herders of the Bale Oromo people in Ethiopia, who were engaged in forest commons management (Rustagi et al. Reference Rustagi, Engel and Kosfeld2010, which Guala cites). The most common behavioral type in the experiments, constituting a bit more than a third of the subjects, were “conditional cooperators” who responded positively to higher contributions by others. Controlling for a large number of other influences on the success of the forest projects, the authors found that groups with more conditional cooperators planted more trees. (See Bowles & Gintis [Reference Bowles and Gintis2011] for more evidence on external validity.)

Second, there is much evidence for costly third-party punishment among societies without formal coercive institutions. Mathew and Boyd (Reference Mathew and Boyd2011) present extensive quantitative data showing that punishment of cowardice and other forms of free-riding plays an important role in warfare among the Turkana, an acephalous African pastoral group. Community members decide whether a violation occurred, and if it has, corporal punishment is administered by the violator's age-mates, not those damaged by the violation. Punishing takes time and effort and may damage valuable social relationships.

Contrary to Guala, punishment has been observed in the simplest foraging societies. Among the Walbiri of Australia, for example, offenses like homicide, physical assault, sacrilege, adultery, and theft were punished by death, wounding with a spear or knife, or attack with a club or boomerang (Meggitt Reference Meggitt1962, pp. 256–59). The local community determined whether the act was an offense, decided on the punishment, nominated the person to carry out the punishment, and appointed the people responsible for ensuring that the punisher does not face retaliation (p. 255).

In some cases, meting out punishment is very costly. Among Aranda foragers of the Central Desert in Australia, wrongdoers were sometimes executed. The elders collectively decided on the fate of the wrongdoer, and assigned a group of young men to carry out the execution. Strehlow (Reference Strehlow and Berndt1970, pp. 117–18) describes two cases in which the violator's relatives did not think the execution was justified, and killed the young men who had carried it out. According to Strehlow, capital punishment of this nature occurred in all Central Australian tribes before colonial administration made them a criminal offense.